Saudi Arabia's AI Data Center Power Constraint Is the Grid, Not the Chips

Power. Saudi Arabia’s AI data center power constraint has a number attached to it: the kingdom operated 467 megawatts of data center load in the first quarter of 2026, on the Ministry of Communications and Information Technology’s (MCIT) own figures [S2], against HUMAIN’s pipeline of 6.6 gigawatts by 2034 [S7]. That is a gap of roughly 14 times, and no export licence closes it.

The chip story is the one that gets written — export controls, Country Group A:5, GB300 allocations, the 600,000-GPU headline. All of it is visible, adversarial and easy to report. The electricity story is none of those things. It is a queue of substations, 380kV feeders, gas turbines with multi-year lead times and grid connection agreements that no Saudi AI project has yet made public. Chips are a procurement problem, solvable with money and diplomacy. Electricity is a construction problem, solvable only with time.

This piece makes one argument: grid interconnection, not silicon, sets the timeline for Saudi AI capacity. The accelerator counts, the model-layer partnership memoranda and the gigawatt announcements are all downstream of when the electrons arrive at a specific parcel of land.

The short version. A data center is an electricity load with servers attached. Announcing gigawatts costs nothing; energising them requires generation, transmission, substations and a signed connection. Saudi Arabia has abundant fuel and cheap tariffs but a national grid built for air conditioning, not for 100kW racks. Until a project publishes a dated energisation record, its megawatts are a plan, not a capacity.

Last verified: 31 July 2026.

What Happened: Saudi Arabia’s AI Data Center Power in Four Numbers

467MW operating. MCIT reported on 5 May 2026 that operational data center capacity had risen to 467MW in Q1 2026, from 440MW at end-2025 and 68MW in 2021 — a near-sixfold increase in four years across more than 60 facilities [S2]. The 440MW figure is the one most widely quoted; it is a 2025 number, not a current one. Both figures come from the same ministry, and our facility-by-facility ledger of what is actually running breaks the total down site by site.

6.6GW promised. HUMAIN targets 6.6GW of data center capacity by 2034, backed by 211 land plots and claimed access to 14GW of power [S7]. A parallel figure of 6GW by 2034 appears in the AMD–Cisco–HUMAIN joint venture framing of November 2025 [S8]. Both are live; they are not the same claim. The 6GW figure describes the compute programme the chip partners signed up to, the 6.6GW figure HUMAIN’s own infrastructure pipeline. Against 467MW, the ratio is 12.8× (6GW) or 14.1× (6.6GW).

44th, and second. Saudi Arabia ranks 44th on the International Data Center Authority’s 2026 Global Digital Readiness Index [S1]. In the same period MCIT publicised a Bloomberg analysis placing the kingdom second globally for data center market attractiveness [S2]. The two are differently scoped rather than contradictory: attractiveness measures the inputs an investor wants, readiness the infrastructure in place. The revealing detail sits inside the flattering ranking — MCIT reports that power availability and land enablement account for 58% of market attractiveness [S2]. Saudi Arabia’s headline advantage is, on the ministry’s own numbers, mostly an electricity advantage.

Zero documented energisations. Madar Strategy audited 13 Saudi and Emirati AI data center projects totalling roughly 4,000MW of announcements and published the result on 22 July 2026: 0MW confirmed energised with public documentation [S3]. Only one of seven headline projects explicitly defines its IT load. None publishes firm connection megawatts, a connection point, or a project-specific energisation date.

For the industry-level profile of the sector, see our Saudi Arabia data centers encyclopedia entry and the data center sector overview; this analysis covers the power constraint specifically, not the market. For the company, see HUMAIN and its AI infrastructure pipeline.

Is Power or Chips the Real Constraint on AI in Saudi Arabia?

Power — and the person who says so most plainly runs the body that ranks digital infrastructure for a living. Mehdi Paryavi, founder and chief executive of the International Data Center Authority (IDCA), told Asharq Al-Awsat in an interview published 21 July 2026 that “Energy, not chips or capital, is the biggest constraint on AI growth, followed by workforce availability and public policy” [S1]. He put Saudi operating data center load at 440MW and the UAE’s at 340MW, together close to 80% of the Middle East’s roughly 1GW total.

That hierarchy — energy, then workforce, then policy, with chips and capital below all three — inverts the way Saudi AI is usually covered. It is also consistent with the observable behaviour of the buildout. Saudi Arabia does not lack authorised chips: HUMAIN holds a US authorisation for the equivalent of up to 35,000 GB300-class accelerators — a case-by-case licence rather than the blanket clearance Washington extended to the Emirates in July 2026 — and confirmed phase-one hardware stands at 18,000 Nvidia GB300 systems. What it lacks is places to plug them in. Our separate analysis of Nvidia GPUs and Saudi chip export controls covers the licensing question in full; the point here is that the licensing question stopped being the binding one some time ago.

The clearest evidence is HUMAIN’s own procurement. On 4 May 2026 the company tendered the enabling infrastructure for its 6GW campus on a 24-square-kilometre site at Al-Saad in east Riyadh, to be delivered in two phases across six plots of 1GW each [S4]. The scope was not compute. It was a 380kV/132kV/33kV electrical distribution network, two substations of 500MVA and 200MVA, and a 2,000MVA bulk supply point, plus water, sewage, stormwater, roads and fibre [S4]. Bids closed 8 May on an early-contractor-involvement basis, meaning that as of mid-2026 the flagship campus was still selecting a partner to design its power infrastructure.

A company that believed chips were its constraint would not spend 2026 tendering switchgear.

How Much Electricity Does a 1GW AI Data Center Need?

A gigawatt of continuous data center load consumes roughly 7.9 terawatt-hours a year at a 90% load factor. The US Energy Information Administration puts average annual American household electricity use at about 10,500 kWh [S17], which makes 1GW equivalent to around 750,000 US homes — and, at typical reactor availability, the continuous output of one large nuclear reactor.

The table below converts the gigawatt claims in circulation into units that can be checked. All conversions assume an 8,760-hour year at a 90% load factor; the Saudi capacity share uses the end-2024 installed base of 121.4GW.

Two things fall out of the arithmetic. First, the entire existing Saudi data center estate — every facility, every operator, sixty-plus sites — draws less than half the power of one large reactor. Second, the 6.6GW pipeline would consume around 52TWh a year. Saudi Arabia delivered a record 160.5TWh to grid customers in the first half of 2025, an annualised rate near 321TWh. On that basis the 2034 pipeline represents roughly 16% of everything the national grid currently delivers — a share no country has added for a single industry inside a decade without building generation specifically for it.

This is why the gigawatt framing matters more than the GPU framing. A 600,000-GPU target is a purchase order. A 6.6GW target is a national infrastructure programme.

Does Saudi Arabia Have Enough Power for 6 Gigawatts of Data Centers?

In aggregate, comfortably. In the specific, unproven.

The three capacity numbers, and which one matters

Three different figures circulate for Saudi generating capacity, and they measure different things:

Against 121.4GW, 6.6GW is 5.4%. Against the grid-connected 92.2GW it is 7.2%. Either way the national fleet is not the binding number. Summer peak demand exceeds 65GW, roughly double the winter baseline, with air conditioning driving about 70% of that peak — so the system already carries a large seasonal swing and has headroom in absolute terms. Our Saudi electricity consumption page tracks the demand series in detail.

What matters instead is firm capacity at a specific location, with a signed connection. A 1GW campus at Al-Saad does not draw on the national average; it draws on the Riyadh transmission system, through a bulk supply point that has to be built, from generation that has to be dispatchable at 3am in August when residential cooling is already at peak. The relevant scarcity is not gigawatts in the country. It is megavolt-amperes at the fence line.

This is the distinction that Madar Strategy’s audit operationalises, and it is the reason the audit found zero documented energisations across 4,000MW of announcements [S3]. Eleven of thirteen cases had no project-specific power evidence at all; one named a utility; one referenced a dedicated grid asset that was not energised. None established a firm connection.

How Long Does It Take to Energise a New AI Data Center?

Longer than it takes to build one — and the gap is widening. Construction of a greenfield facility runs 18 to 24 months from groundbreaking, with a further three to six months of testing and commissioning. Power delivery in constrained markets averages around four years, and reaches seven years in Northern Virginia for a 100MW request [S20]. Large power transformers carry two-to-four-year lead times; medium-voltage switchgear and generators have multi-year backlogs [S20]. As one industry assessment puts it, the building goes up faster than the electrons can be secured to fill it.

Saudi Arabia is not the United States. It has a single buyer, a state-directed grid, no local-opposition politics and a sovereign willing to prioritise. Those advantages are real and they compress the queue. They do not compress transformer manufacturing, and they do not remove the sequence. Every project must pass the same six gates:

Gates one through three are visible from a press release. Gates four through six are the ones that decide whether a gigawatt exists, and they are precisely the gates on which the Saudi AI programme publishes nothing. That is not necessarily concealment — connection terms are commercially sensitive everywhere. But it means that as of 28 July 2026, no external party can verify that a single megawatt of announced Saudi AI capacity has been energised.

The Near-Term Power Answer Is Gas Turbines, Not Renewables

The generation Saudi Arabia is actually procuring to meet this load is gas.

On 13–14 July 2026 the Saudi Power Procurement Company (SPPC), the kingdom’s single buyer of electricity, opened qualification for a third round of combined-cycle gas turbine (CCGT) independent power projects, tendered under Ministry of Energy supervision [S5][S6]. Each plant is developed on a build-own-operate basis, with the winning consortium taking 100% equity in a project company, and each is designed with provision for carbon capture readiness so the equipment can be retrofitted later [S5].

Resolving the 1.8GW confusion

Trade coverage of the round has been reported in places as a 1.8GW programme. That is a misreading of the unit. 1,800MW is the capacity of each individual project, not the round total [S5][S6]. The first round comprised Taiba 1, Taiba 2, Qassim 1 and Qassim 2 — four plants at 1,800MW each, 7,200MW combined. The second comprised Rumah 1, Rumah 2, Nairiyah 1 and Nairiyah 2, again 7,200MW combined. On the same four-by-1,800MW structure, round three is a 7.2GW procurement, not a 1.8GW one. Anyone modelling Saudi supply additions from the 1.8GW figure understates the round by a factor of four.

Set that against the renewable position. Saudi Arabia targets 130GW of renewable capacity by 2030 and 50% of generation from renewables [S13]. The 2024 starting point was close to the opposite. On Ember’s generation data, natural gas supplied 63.3% of Saudi electricity in 2024 and other fossil sources, overwhelmingly oil, 34.5% — leaving renewables at 2.2% and fossil fuels at 97.8% of the mix [S21]. Power-market analysis puts the same year at 93.2% thermal against 6.8% renewables [S14], most likely because the two series count captive and off-grid solar differently. The generation-basis figure is the conservative one and the more relevant to a grid-connected load.

Whichever series is used, the conclusion holds: the electricity that will run Saudi AI data centers in 2027 and 2028 will be burned, not harvested. The 2026 renewable energy sprint and the 130GW renewable capacity target are real programmes, but the renewable energy gap between installed and targeted capacity remains among the widest in the Vision 2030 KPI set. The gas expansion programme, anchored on the Jafurah gas field, is the supply story that matches the AI timeline.

The narrative cost is straightforward. An AI buildout marketed alongside the renewable energy sector is, in the near term, four more combined-cycle gas plants per procurement round — carbon-capture-ready rather than carbon-captured.

Cheap Subsidised Electricity Is the Product Saudi Arabia Is Selling

Strip away the sovereign-AI language and the commercial proposition is a tariff.

Saudi Arabia’s regulated industrial electricity rate is SAR0.18 per kWh (about $0.048), with commercial users at SAR0.20 per kWh up to 6,000 kWh a month and SAR0.30–0.32 above it, under the schedule administered by the Saudi Electricity Regulatory Authority (SERA) [S22]. Measured across all business consumption, Saudi Arabia’s price averaged $0.074 per kWh in December 2025 — roughly 45% of the world average, with households at $0.053 [S23]. HUMAIN has told prospective equity partners that its facilities benefit from subsidised electricity, and has framed the resulting cost base as up to 30% below US rates [S7].

For a 1GW facility drawing 7.9TWh a year, a 5-cent delta against a 10-cent market is roughly $395m a year in avoided energy cost. That is the entire arbitrage. It is also a fiscal transfer: subsidised power sold to a PIF-owned company is a claim on the same budget that funds the rest of the transformation, and it scales linearly with every gigawatt added.

Saudi Arabia’s second-place ranking for market attractiveness is, more than anything else, a ranking of its electricity system — which is why 44th place on infrastructure readiness [S1] reads as a delivery warning rather than a contradiction. The kingdom is selling something it demonstrably has. The question is how fast it can wire it.

Cooling and Water Are the Second-Order Power Problem

Cooling in a desert is an electricity problem before it is a water problem, and both are power problems in the end.

Ambient temperatures across the Eastern Province and Riyadh push mechanical cooling loads well above the levels assumed in temperate-climate designs, which is why Saudi projects are specified around direct liquid cooling and dry or hybrid systems rather than evaporative towers. SDAIA’s 480MW Hexagon facility in Riyadh — a $2.7bn (SAR10.1bn) government data center whose foundation stone was laid in January 2026, certified Tier IV and built to LEED Gold — uses direct liquid cooling and hybrid systems specifically to hold power usage effectiveness down [S18]. DataVolt’s 1.5GW campus at Oxagon in NEOM, a $5bn first phase with a 300MW opening tranche targeted for 2028, is designed to run at net zero on renewable supply [S19].

The water arithmetic is unforgiving. A 1GW facility in Saudi Arabia is estimated to require 1.5 to 5 million cubic metres of water a year. Gulf operators target water usage effectiveness below 0.2 litres per kWh, well under global averages, and liquid cooling can cut cooling water use by up to 90% [S15]. But roughly 70% of Saudi drinking water comes from desalination — itself among the most electricity-intensive processes in the kingdom, as our desalination capacity page sets out. Water saved is electricity saved; water consumed is electricity consumed twice.

Abdalftah Hamed Ali and Mohammad Abu Hawash of the Middle East Council on Global Affairs argued in April 2026 that Saudi data center power demand is on a 29% compound annual growth rate, and that “the Gulf cannot scale AI sustainably unless it treats water efficiency and summer peak resilience as first-order design constraints, not downstream externalities” [S15]. Summer peak resilience is the phrase that matters: an AI campus is a baseload draw layered on top of a system whose peak is already set by air conditioning. KAPSARC, the kingdom’s own energy research centre, reached a parallel conclusion in its December 2025 paper with ICAIRE — AI-ready facilities are defined by dense servers and liquid cooling, and their economics are most sensitive to utilisation and hardware efficiency [S16].

The Slippage Is Already Visible in the Public Record

If power were not the binding constraint, the delivery record would be cleaner than it is.

The stc–HUMAIN joint venture has been extended rather than signed. stc and HUMAIN signed a memorandum of understanding on 18 December 2025 to build up to 1GW of AI data center capacity through stc’s subsidiary Center3, with HUMAIN taking 51% and stc 49%. Six months later, on 18 June 2026, a Tadawul filing disclosed a six-month extension rather than an executed joint venture, citing regulatory and operational requirements still outstanding [S9]. The initial phase remains “up to 250 megawatts, subject to customer commitments” — a demand condition attached to the smallest tranche of the smallest flagship project.

HUMAIN’s twin campuses have not been confirmed live. The company’s 100MW facilities in Riyadh and Dammam were expected to begin operations in Q2 2026. As of 28 July 2026 no announcement or reporting confirms either is operational; retrievable sources still use forward-looking language. Absence of an announcement is not proof of delay. But a state programme that publicises foundation stones does not usually go quiet about switch-ons.

No project has published an energisation date. Across 4,000MW of Saudi and Emirati announcements, zero energisation records exist in public [S3]. The strongest case in the region — a 100MW facility in Ajman with defined IT load, completed steel and a named utility — still has none, and has moved its completion target from Q3 2025 to December 2026 [S3].

The honest counter-argument deserves stating: none of this proves failure. Grid connection agreements are commercially confidential in every market, Saudi utilities are not obliged to publish them, and the Saudi Electricity Company has a capital programme running to hundreds of billions of riyals and a demonstrated ability to build fast when directed. The kingdom’s national AI strategy is coordinated through SDAIA with a directness most jurisdictions cannot match. 2027 may well produce a run of energisations that makes this section look pessimistic. The claim here is narrower and harder to dispute: on the public record as it stands, the megawatt is unverified and the gigawatt is a plan.

War Risk Has Become a Data Center Power Risk

The 2026 conflict converted an abstract siting question into a physical one.

The OECD warned on 7 June 2026 that prolonged Middle East conflict and the effective closure of the Strait of Hormuz could delay or halt Gulf AI ambitions, noting that large-scale AI infrastructure projects are tied to sovereign wealth funds and state-backed capital, that higher energy prices raise data center operating costs, and that disruption to helium exports feeds back into semiconductor production. Regional data center growth had been forecast at 63% annually before the conflict. The warning cited March 2026 drone attacks on cloud facilities in the UAE and Bahrain and named Saudi Arabia’s $2.7bn Hexagon facility among the assets at risk [S10].

It escalated from warning to event. On 21 July 2026 Iran’s Islamic Revolutionary Guard Corps claimed cruise-missile strikes on Amazon data center infrastructure in Bahrain. Bloomberg reported on 28 July that satellite imagery confirmed damage to two Amazon facilities, at Zallaq and Askar; Amazon had already migrated customers to unaffected sites [S11].

Then, on 27 July 2026, Saudi air defences intercepted drones targeting oil facilities in the Eastern Province and Riyadh. The Ministry of Defence attributed the attack to Iran-backed militias operating from Iraqi territory; spokesman Major General Turki Al Malki called them terrorist attacks on energy infrastructure and said the kingdom “reserves the right to defend itself and respond at the appropriate time and place.” All drones were intercepted and no damage was reported. The Islamic Resistance in Iraq denied launching them from Iraqi territory, and Iraq’s government ordered an investigation [S12].

The targets were oil facilities, not data centers. But HUMAIN’s Dammam campus sits in the Eastern Province, inside the same air-defence envelope. Three consequences follow for the power thesis: higher insurance and hardening costs for every megawatt built east of Riyadh; a stronger case for siting AI load inland or at NEOM, which lengthens transmission runs; and grid resilience — redundant feeders, on-site generation, black-start capability — becoming a design requirement rather than an option, adding cost and months to exactly the gates that are already slowest.

Why This Matters for Vision 2030

Vision 2030’s AI ambition is stated in compute terms: third-largest AI provider globally, 7% of world training and inference workloads by 2030. Those are power targets wearing compute clothing. Seven per cent of global AI workload cannot be served from 467MW, and the difference between the two is not chips.

Three consequences follow. Fiscally, subsidised electricity sold to a PIF-owned company converts an energy endowment into an off-balance-sheet subsidy that scales with every gigawatt, competing with the domestic demand growth that already pushes summer peak past 65GW. Structurally, the programme’s dependence on gas rather than renewables sits in direct tension with the 130GW renewable target, since every CCGT round adds fossil baseload with a 25-year life. Reputationally, the gap between announced and energised megawatts is the most falsifiable claim in the Year of AI narrative. Compute promises are hard to audit. Energisation dates are not — which is why they will eventually be the metric used to grade the programme.

Risks, Contradictions and Open Questions

We do not know whether HUMAIN’s Riyadh and Dammam campuses are live. No confirmation exists either way as of 28 July 2026. This is the highest-value unresolved fact on the beat.

No connection agreement has been published. Without firm connection megawatts, a connection point and an energisation date, external capacity estimates for Saudi AI infrastructure are inference from press releases — for bullish and bearish readings alike.

The 6GW and 6.6GW targets have never been reconciled publicly. They may describe the same programme measured differently, or two overlapping pipelines. HUMAIN has not clarified.

Installed-capacity figures disagree by nearly 30GW. 121.4GW (installed, EIA-derived, end-2024), 100.6GW (installed base, 2025) and 92.2GW (grid-connected, 2024) are all in circulation. The differences are definitional, but they change every percentage built on them.

The renewables share of 2024 generation is disputed — 2.2% on Ember’s generation data against 6.8% on power-market data [S14][S21]. Neither figure changes the conclusion that near-term AI load will be gas-fired.

Cooling water demand is estimated, not measured. The 1.5–5 million cubic metre range for a 1GW facility spans more than a factor of three, and no Saudi operator publishes actual water usage effectiveness.

The counter-case is genuine. A single-buyer market, a sovereign that can direct SPPC and the Saudi Electricity Company, cheap fuel, empty land and no permitting opposition are advantages the US and European markets lack. If any jurisdiction can compress a four-year energisation queue, it is this one. The thesis is that the sequence still binds — not that it cannot be shortened.

What to Watch Next

• SPPC third-round CCGT qualification results. Watch for the named sites and the confirmed round total. Four projects of 1,800MW settles the 7.2GW reading.
• The stc–HUMAIN MoU expiry, on or about 18 December 2026. A second six-month extension, or a lapse, would be a stronger delivery signal than any announcement. Execution with a committed 250MW anchor tenant would be the opposite.
• Any dated energisation record. The first Saudi AI project to publish a grid connection date, connection point and firm megawatts resets this analysis. None has yet.
• HUMAIN Riyadh and Dammam confirmations. Q2 2026 has passed. A live-service announcement, or continued silence into Q4, is diagnostic.
• Q3 2026 MCIT capacity update. The series ran 68MW (2021) → 440MW (2025) → 467MW (Q1 2026). The next print shows whether the curve is bending upward or flattening.
• Hexagon and DataVolt milestones. Hexagon’s 480MW and DataVolt’s first 300MW at Oxagon are the largest near-term additions; both are construction-stage, neither energised.
• Eastern Province security posture. Further strikes near Dammam would move siting economics, insurance and hardening costs materially.

Related Vision 2030 Context

• Saudi Arabia data centers: industry growth and investment — the sector’s full profile, capacity history and operator landscape.
• HUMAIN: the PIF-owned AI company — ownership, mandate and partnerships.
• HUMAIN’s AI infrastructure pipeline — the 600,000-GPU and 6.6GW programme in detail.
• Saudi Electricity Company — the utility that has to deliver the connections.
• Saudi electricity consumption — demand, generation and tariff reform.
• Saudi Arabia’s 130GW renewable capacity target — what is installed against what is promised.
• Saudi AI cloud and compute infrastructure — hyperscaler regions and sovereign compute.
• Nvidia GPUs, Saudi AI and export controls — the chip-access question this piece deliberately sets aside.

Sources

• [S1] Arab News, Head of International Data Center Authority to Asharq Al-Awsat: Saudi Arabia Leads AI Race, interview report, 21 July 2026. https://www.arabnews.com/node/2651715/business-economy
• [S2] Ministry of Communications and Information Technology (Saudi Arabia), Saudi Arabia Ranks Second Globally in Data Center Market Attractiveness, official statement, 5 May 2026. https://mcit.gov.sa/en/node/249183
• [S3] Madar Strategy, Headline Megawatts: An Evidence Audit of Saudi and UAE AI Data Centers, research note, 22 July 2026. https://www.madarstrategy.com/research/headline-megawatts
• [S4] MEED, Humain Tenders Infrastructure for 6GW Data Centre Campus, trade report, 4 May 2026. https://guest.meed.com/humain-tenders-6gw-data-centre-campus-infrastructure/
• [S5] SaudiGulf Projects, Principal Buyer Launches Qualification for Third Round of CCGT Independent Power Producer Projects in Saudi Arabia, trade report, July 2026. https://www.saudigulfprojects.com/2026/07/sppc-third-round-ccgt-ipp-projects-saudi-arabia/
• [S6] SolarQuarter, Saudi Principal Buyer Opens Qualification for Third Round of CCGT IPP Projects, trade report, 14 July 2026. https://solarquarter.com/2026/07/14/saudi-principal-buyer-opens-qualification-for-third-round-of-ccgt-ipp-projects/
• [S7] Data Center Dynamics, Saudi Arabia’s AI Co. Humain Looking for US Data Center Equity Partner, Targets 6.6GW by 2034 with Subsidized Electricity, trade report, 2026. https://www.datacenterdynamics.com/en/news/saudi-arabias-ai-co-humain-looking-for-us-data-center-equity-partner-targets-66gw-by-2034-with-subsidized-electricity/
• [S8] AMD, AMD, Cisco and HUMAIN to Form Joint Venture to Deliver World-Leading AI Infrastructure, press release, 19 November 2025. https://www.amd.com/en/newsroom/press-releases/2025-11-19-amd-cisco-and-humain-to-form-joint-venture.html
• [S9] Argaam, stc Extends HUMAIN AI Data Center MoU by 6 Months, market report on Tadawul filing, 18 June 2026. https://www.argaam.com/en/article/articledetail/id/1913648
• [S10] AGBI, Hormuz Disruption Threatens AI Hub Ambitions, OECD Warns, news report, 7 June 2026. https://www.agbi.com/ai/2026/06/hormuz-disruption-threatens-ai-hub-ambitions-oecd-warns/
• [S11] Bloomberg, Amazon Data Centers Hit in Iran Strikes, Satellite Images Show, news report, 28 July 2026. https://www.bloomberg.com/news/articles/2026-07-28/amazon-data-centers-hit-in-iran-strikes-satellite-images-show
• [S12] The National, Saudi Arabia ‘Reserves Right to Respond’ After Drone Attacks From Iraq, news report, 27 July 2026. https://www.thenationalnews.com/news/gulf/2026/07/27/saudi-arabia-drone-attacks-iraq/
• [S13] International Trade Administration, US Department of Commerce, Saudi Arabia — Power, country commercial guide, updated 13 May 2026. https://www.trade.gov/country-commercial-guides/saudi-arabia-power
• [S14] Mordor Intelligence, Saudi Arabia Power Market Size and Share Analysis, market report, 2026. https://www.mordorintelligence.com/industry-reports/saudi-arabia-power-market
• [S15] Middle East Council on Global Affairs, Abdalftah Hamed Ali and Mohammad Abu Hawash, Desert Bytes: Why Gulf AI Ambitions Must Align With Energy and Water Realities, policy publication, 9 April 2026. https://mecouncil.org/publication/gulf-ai-data-centers-water-energy/
• [S16] KAPSARC and ICAIRE, AI and Energy: The Future of Data Centers in Saudi Arabia, discussion paper KS-2025-DP69, December 2025. https://www.kapsarc.org/our-offerings/publications/ai-and-energy-the-future-of-data-centers-in-saudi-arabia/
• [S17] US Energy Information Administration, How Much Electricity Does an American Home Use?, FAQ, 2026. https://www.eia.gov/tools/faqs/faq.php?id=97&t=3
• [S18] w.media, SDAIA Lays Foundation Stone for 480MW Hexagon Government Data Center in Riyadh, trade report, 5 January 2026. https://w.media/sdaia-lays-foundation-stone-for-480mw-hexagon-government-data-center-in-riyadh/
• [S19] ESG Today, Saudi’s NEOM, DataVolt to Build $5 Billion Net Zero AI Data Center, news report, 2025. https://www.esgtoday.com/saudis-neom-datavolt-to-build-5-billion-net-zero-ai-data-center/
• [S20] Construction Owners, How Long It Actually Takes to Power a Data Center in 2026: A US Market-by-Market Reality Check, industry analysis, 2026. https://www.constructionowners.com/insights/how-long-it-actually-takes-to-power-a-data-center-in-2026-a-u-s-market-by-market-reality-check
• [S21] Ember, Saudi Arabia Electricity Data, 2024 generation series. https://ember-energy.org/countries-and-regions/saudi-arabia/ — Ember’s site returned an access error to automated retrieval; the 2024 shares quoted here (gas 63.3%, other fossil 34.5%, renewables 2.2%) are as reported from Ember’s dataset in secondary coverage.
• [S22] Saudi Gazette, Revised Residential and Commercial Electricity Tariffs to Remain Same, news report on the Saudi Electricity Regulatory Authority schedule. https://saudigazette.com.sa/article/652853
• [S23] GlobalPetrolPrices, Saudi Arabia Electricity Prices, data series, December 2025. https://www.globalpetrolprices.com/Saudi-Arabia/electricity_prices/